BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Artificial Intelligence Research Group Talks (Comp
 uter Laboratory)
SUMMARY:Predicting Mortality and Length of Stay with Patie
 nt Graph Representation Learning - Emma Rocheteau
DTSTART;TZID=Europe/London:20201020T131500
DTEND;TZID=Europe/London:20201020T141500
UID:TALK152929AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/152929
DESCRIPTION:"Join us on Zoom":https://zoom.us/j/99166955895?pw
 d=SzI0M3pMVEkvNmw3Q0dqNDVRalZvdz09\n\nRecent work 
 on predicting patient outcomes in the Intensive Ca
 re Unit (ICU) has focused heavily on the physiolog
 ical time series data\, largely ignoring sparse da
 ta such as diagnoses and medications. When they ar
 e included\, they are usually concatenated in the 
 late stages of a model\, which may struggle to lea
 rn from rarer disease patterns. Instead\, we propo
 se a strategy to exploit diagnoses as relational i
 nformation by connecting similar patients in a gra
 ph. To this end\, we propose LSTM-GNN for patient 
 outcome prediction tasks: a hybrid model combining
  Long Short-Term Memory networks (LSTMs) for extra
 cting temporal features and Graph Neural Networks 
 (GNNs) for extracting the patient neighbourhood in
 formation. We demonstrate that LSTM-GNNs outperfor
 m the LSTM-only baseline on length of stay predict
 ion tasks on the eICU database. More generally\, o
 ur results indicate that exploiting information fr
 om neighbouring patient cases using graph neural n
 etworks is a promising research direction\, yieldi
 ng tangible returns in supervised learning perform
 ance on EHRs.
LOCATION:Zoom
CONTACT:Mateja Jamnik
END:VEVENT
END:VCALENDAR